| |
| """Validate parity between the repo and prepared Hugging Face mirrors. |
| |
| This is a publisher-side check. It compares critical website data, figures, and |
| validator scripts across the local repo, prepared HF Space bundle, prepared HF |
| artifact dataset bundle, and prepared HF model bundle before upload. |
| """ |
|
|
| from __future__ import annotations |
|
|
| import argparse |
| import hashlib |
| import json |
| from datetime import datetime, timezone |
| from pathlib import Path |
|
|
|
|
| ROOT = Path(__file__).resolve().parents[1] |
| DEFAULT_HF_ROOT = ROOT.parent / "hf_publish" |
| DEFAULT_OUTPUT = ROOT / "docs/data/mirror_parity.json" |
| QWEN3_FUTURE_TASK_PROBE_RUN_ID = "xperience10m_qwen3_omni_v6_future_task_probes_a100_20260616T143608Z" |
| QWEN3_RETRIEVAL_TASK_PROBE_RUN_IDS = [ |
| "xperience10m_qwen3_omni_v6_retrieval_task_probes_a100_20260617T175919Z", |
| "xperience10m_qwen3_omni_v6_cross_modal_retrieval_probe_a100_20260618T000000Z", |
| "xperience10m_qwen3_omni_v6_camera_view_sync_mosaic_tile_a100_20260619T0305Z", |
| "xperience10m_qwen3_omni_v6_sensor_target_probes_a100_20260619T000000Z", |
| "xperience10m_qwen3_omni_v6_interaction_text_task15_a100_20260620T010305Z", |
| ] |
| COSMOS3_SUPER_RETRIEVAL_TASK_PROBE_RUN_IDS = [ |
| "xperience10m_cosmos3_super_retrieval_task_probes_a100_textonly_prompatch_v2_20260620", |
| "xperience10m_cosmos3_nano_retrieval_task_probes_a100_patched_textonly_20260621", |
| ] |
| COSMOS3_SUPER_FUTURE_TASK_PROBE_RUN_IDS = [ |
| "xperience10m_cosmos3_super_future_task_probes_a100_textonly_v1_20260620", |
| "xperience10m_cosmos3_nano_future_order_misalignment_patched_textonly_20260621", |
| "xperience10m_cosmos3_nano_current_subtask_object_relevance_patched_textonly_20260621", |
| ] |
| COSMOS3_SUPER_INTERACTION_TEXT_TASK_PROBE_RUN_IDS = [ |
| "xperience10m_cosmos3_super_interaction_text_task15_textonly_v1_20260620T1558Z", |
| "xperience10m_cosmos3_nano_interaction_text_task15_patched_textonly_20260621", |
| ] |
|
|
| DATA_FILES = [ |
| "additional_development_directions.json", |
| "audio_ablation_summary.json", |
| "artifact_index.json", |
| "brand_assets.json", |
| "evidence_contract.json", |
| "evaluation_protocol.json", |
| "figure_index.json", |
| "foundation_model_plan.json", |
| "glossary.json", |
| "live_publication_status.json", |
| "language_versions.json", |
| "modality_atlas.json", |
| "omni_finetune_verified_result.json", |
| "omni_model_comparison.json", |
| "project_brief.json", |
| "project_manifest.json", |
| "project_packet.json", |
| "public_reader_map.json", |
| "project_status.json", |
| "publication_audit.json", |
| "public_surface_qa.json", |
| "qwen3_full_parameter_gates.json", |
| "qwen3_omni_run_lineage.json", |
| "qwen3_v5_v6_comparison.json", |
| "quality_gates.json", |
| "raw_sample_files.json", |
| "rendered_site_check.json", |
| "reproducibility_matrix.json", |
| "research_roadmap.json", |
| "research_roadmap_interactive.json", |
| "research_takeaways.json", |
| "research_direction_extensions.json", |
| "research_directions.json", |
| "scope_claims_audit.json", |
| "single_episode_explorer.json", |
| "source_alignment_audit.json", |
| "summary_metrics.json", |
| "single_episode_task_model_radar.json", |
| "episode128_task_model_radar.json", |
| "three_foundation_pipelines.json", |
| "two_evidence_lines.json", |
| "two_evidence_line_result_summary.json", |
| "task_suite_20.json", |
| "task_suite_enhancement_128.json", |
| "task_surface_integrity.json", |
| "task_walkthroughs.json", |
| "task_method_20_result_matrix.json", |
| "task_method_20_gap_audit.json", |
| "task_method_20_source_audit.json", |
| "tier2_task_suite.json", |
| "unified_task_model_radar.json", |
| "website_integrity.json", |
| "xperience10m_128_episode_feature_index.json", |
| "xperience10m_dataset_card_alignment.json", |
| ] |
|
|
| ASSET_FILES = [ |
| "charts/audio_ablation_delta.svg", |
| "charts/two_evidence_line_map.svg", |
| "charts/single_episode_task_model_radar.svg", |
| "charts/episode128_task_model_radar.svg", |
| "charts/tier2_task_suite.svg", |
| "charts/unified_task_model_radar.svg", |
| "charts/research_direction_coverage.svg", |
| "brand/xperience10m-logo-apple-touch.png", |
| "brand/xperience10m-logo-favicon-32.png", |
| "brand/xperience10m-logo-favicon-64.png", |
| "brand/xperience10m-logo-mark.png", |
| "brand/xperience10m-logo-mark-192.png", |
| "brand/xperience10m-logo-mark-512.png", |
| "brand/xperience10m-logo-social-card.png", |
| "task_suite_infographic.png", |
| "pipeline_diagram.png", |
| "pipeline_diagram.svg", |
| "task_architectures.png", |
| "task_architectures.svg", |
| "foundation-pipelines/spatial-intelligence-pipeline.png", |
| "foundation-pipelines/human-video-world-model-pipeline.png", |
| "foundation-pipelines/vision-language-action-pipeline.png", |
| "foundation-pipelines/README.md", |
| "foundation-pipelines/prompts.md", |
| "foundation-pipelines/source-slides/spatial-intelligence-slide.png", |
| "foundation-pipelines/source-slides/human-video-world-model-slide.png", |
| "foundation-pipelines/source-slides/vision-language-action-slide.png", |
| "foundation-pipelines/source-photos/spatial-intelligence-source.jpg", |
| "foundation-pipelines/source-photos/human-video-world-model-source.jpg", |
| "foundation-pipelines/source-photos/vision-language-action-source.jpg", |
| "modalities/audio.png", |
| "modalities/depth.jpg", |
| "modalities/inertial.png", |
| "modalities/language.png", |
| "modalities/motion_capture.png", |
| "modalities/pose_slam.png", |
| "modalities/video.jpg", |
| "raw-sample-preview/fisheye_cam0_preview.mp4", |
| "raw-sample-preview/fisheye_cam1_preview.mp4", |
| "raw-sample-preview/fisheye_cam2_preview.mp4", |
| "raw-sample-preview/fisheye_cam3_preview.mp4", |
| "raw-sample-preview/stereo_left_preview.mp4", |
| "raw-sample-preview/stereo_right_preview.mp4", |
| ] |
|
|
| SCRIPT_FILES = [ |
| "omni/analyze_qwen3_omni_errors.py", |
| "omni/audit_cosmos3_super_training_contract.py", |
| "omni/build_omni_model_comparison.py", |
| "omni/build_qwen3_full_parameter_gate_summary.py", |
| "omni/build_128_episode_feature_index.py", |
| "omni/collect_qwen3_future_task_probe_results.sh", |
| "omni/collect_qwen3_retrieval_task_probe_results.sh", |
| "omni/collect_cosmos3_super_retrieval_task_probe_results.sh", |
| "omni/collect_cosmos3_super_future_task_probe_results.sh", |
| "omni/eval_cosmos3_super_interaction_text_task.py", |
| "omni/collect_qwen3_v4_release_artifacts.py", |
| "omni/defer_cosmos3_super_after_qwen_v4.sh", |
| "omni/defer_qwen3_retrieval_after_order_sync.sh", |
| "omni/defer_qwen3_fullparam_after_verified_qwen.sh", |
| "omni/eval_qwen3_omni_future_task_probes.py", |
| "omni/eval_qwen3_omni_retrieval_task_probes.py", |
| "omni/eval_cosmos3_super_future_task_probes.py", |
| "omni/eval_cosmos3_super_retrieval_task_probes.py", |
| "omni/export_cosmos3_camera_pose_targets.py", |
| "omni/launch_all_task_model_scoring_when_free.sh", |
| "omni/merge_qwen3_omni_future_task_probe_shards.py", |
| "omni/merge_qwen3_omni_retrieval_task_probe_shards.py", |
| "omni/merge_cosmos3_super_future_task_probe_shards.py", |
| "omni/merge_cosmos3_super_interaction_text_task_shards.py", |
| "omni/pack_cosmos3_super_action_batch.py", |
| "omni/prepare_cosmos3_super_lora_hf_package.py", |
| "omni/prepare_qwen3_lora_hf_package.py", |
| "omni/patch_qwen3_omni_video_features.py", |
| "omni/probe_cosmos3_super_training_readiness.py", |
| "omni/run_private_gpu_qwen3_v6_repro_smoke.sh", |
| "omni/run_qwen3_omni_future_task_probes_sharded.sh", |
| "omni/run_qwen3_omni_retrieval_task_probes_sharded.sh", |
| "omni/run_cosmos3_super_future_task_probes_sharded.sh", |
| "omni/run_cosmos3_super_retrieval_task_probes_sharded.sh", |
| "omni/run_cosmos3_super_interaction_text_task_sharded.sh", |
| "omni/score_existing_model_output_task_probes.py", |
| "omni/score_model_output_probes.py", |
| "omni/run_128_task_baselines.py", |
| "omni/build_task_suite_enhancement_128.py", |
| "omni/run_cosmos3_super_forward_dynamics_lora.sh", |
| "omni/train_cosmos3_super_forward_dynamics_lora.py", |
| "audio_ablation_and_raw_upgrade.py", |
| "build_artifact_index.py", |
| "build_brand_assets.py", |
| "build_evaluation_protocol.py", |
| "build_figure_index.py", |
| "build_quality_gates.py", |
| "build_qwen3_omni_run_lineage.py", |
| "build_public_surface_qa.py", |
| "build_rendered_site_check.py", |
| "build_interactive_research_roadmap.py", |
| "build_multilingual_public_readmes.py", |
| "render_foundation_pipeline_diagrams.py", |
| "build_single_episode_explorer.py", |
| "build_task_method_20_gap_audit.py", |
| "build_research_takeaways.py", |
| "export_modality_atlas_assets.py", |
| "generate_visualizations.py", |
| "render_overview_figures.py", |
| "render_task_suite_infographic.py", |
| "research_direction_taxonomy.py", |
| "task_display.py", |
| "task_walkthroughs.py", |
| "build_unified_task_suite.py", |
| "build_unified_task_model_radar.py", |
| "single_episode_diagnostics.py", |
| "verify_live_publication.py", |
| "validate_mirror_parity.py", |
| "validate_publication_package.py", |
| "validate_task_method_matrix_sources.py", |
| "validate_scope_claims.py", |
| "validate_source_alignment.py", |
| "validate_task_surface.py", |
| "validate_website_integrity.py", |
| "sync_hf_publish_mirrors.py", |
| "tier2_task_suite.py", |
| "publish_hf_bundles.py", |
| ] |
|
|
| WEBSITE_FILES = [ |
| "404.html", |
| "apple-touch-icon.png", |
| "favicon.png", |
| "index.html", |
| "research_roadmap.html", |
| "single_episode_explorer.html", |
| "site.webmanifest", |
| ] |
|
|
| RESULT_FILES = [ |
| "audio_ablation/AUDIO_ABLATION_SUMMARY.md", |
| "audio_ablation/audio_ablation_metrics.csv", |
| "audio_ablation/audio_ablation_summary.json", |
| "audio_ablation/audio_delta_summary.csv", |
| "audio_ablation/raw_logmel_fisheye_cam0_sr16000_mels64_fft512_hop160.npz", |
| "single_episode_diagnostics/provenance.json", |
| "single_episode_diagnostics/README.md", |
| "single_episode_diagnostics/modality_ablation/ablation_metrics.csv", |
| "single_episode_diagnostics/modality_ablation/ablation_summary.json", |
| "single_episode_diagnostics/object_labels/object_vocab.json", |
| "single_episode_diagnostics/object_labels/window_object_labels.csv", |
| "single_episode_diagnostics/timeline_overlay/timeline_overlay.csv", |
| "single_episode_diagnostics/alignment_stress/alignment_shift_metrics.csv", |
| "single_episode_diagnostics/alignment_stress/alignment_stress_summary.json", |
| "omni_finetune/verified_public/xperience10m_qwen3_omni_128ep_96train_16val_16test_valmon_20260605_eval/analysis/ERROR_ANALYSIS.md", |
| "omni_finetune/verified_public/xperience10m_qwen3_omni_128ep_96train_16val_16test_valmon_20260605_eval/analysis/error_analysis_summary.json", |
| "omni_finetune/verified_public/xperience10m_qwen3_omni_128ep_96train_16val_16test_valmon_20260605_eval/analysis/episode_error_analysis.csv", |
| "omni_finetune/verified_public/xperience10m_qwen3_omni_128ep_96train_16val_16test_valmon_20260605_eval/analysis/action_family_error_analysis.csv", |
| "omni_finetune/verified_public/xperience10m_qwen3_omni_128ep_96train_16val_16test_valmon_20260605_eval/analysis/train_seen_error_analysis.csv", |
| "omni_finetune/verified_public/xperience10m_qwen3_omni_128ep_96train_16val_16test_valmon_20260605_eval/analysis/missing_modality_error_analysis.csv", |
| "omni_finetune/verified_public/xperience10m_qwen3_omni_128ep_96train_16val_16test_valmon_20260605_eval/analysis/object_category_error_analysis.csv", |
| "omni_finetune/multi_episode_128_task_baselines/BASELINE_ALIGNMENT_REPORT.md", |
| "omni_finetune/multi_episode_128_task_baselines/summary_report.json", |
| "omni_finetune/multi_episode_128_task_baselines/task_metrics.csv", |
| "omni_finetune/task_suite_enhancement_128_v1_20260608/ENHANCEMENT_REPORT.md", |
| "omni_finetune/task_suite_enhancement_128_v1_20260608/dense_window_scenarios.csv", |
| "omni_finetune/task_suite_enhancement_128_v1_20260608/enhancement_plan.json", |
| "omni_finetune/task_suite_enhancement_128_v1_20260608/experiment_backlog.json", |
| "omni_finetune/task_suite_enhancement_128_v1_20260608/hierarchical_target_contract.json", |
| "omni_finetune/task_suite_enhancement_128_v1_20260608/qwen_action_family_error_summary.csv", |
| "omni_finetune/task_suite_enhancement_128_v1_20260608/task_bottlenecks.csv", |
| "omni_finetune/OMNI_MODEL_COMPARISON.md", |
| "omni_finetune/QWEN3_FULL_PARAMETER_GATES_20260609.md", |
| "omni_finetune/QWEN3_V5_V6_COMPARISON_20260614.md", |
| "omni_finetune/xperience10m_qwen3_omni_128ep_fullparam_smoke_preemptible_8gpu_20260609/fullparam_feasibility_summary.json", |
| "omni_finetune/xperience10m_qwen3_omni_128ep_fullparam_smoke_preemptible_8gpu_20260609/progress.jsonl", |
| "omni_finetune/xperience10m_qwen3_omni_128ep_fullparam_smoke_preemptible_8gpu_20260609/training_metadata.json", |
| "omni_finetune/xperience10m_qwen3_omni_128ep_fullparam_smoke_preemptible_8gpu_20260609/RUN_REPORT.md", |
| "omni_finetune/xperience10m_qwen3_omni_128ep_fullparam_shorttrain8_preemptible_8gpu_20260609/fullparam_shorttrain8_summary.json", |
| "omni_finetune/xperience10m_qwen3_omni_128ep_fullparam_shorttrain8_preemptible_8gpu_20260609/progress.jsonl", |
| "omni_finetune/xperience10m_qwen3_omni_128ep_fullparam_shorttrain8_preemptible_8gpu_20260609/launch_status.jsonl", |
| "omni_finetune/xperience10m_qwen3_omni_128ep_fullparam_pilot32_preemptible_8gpu_20260609/fullparam_pilot32_summary.json", |
| "omni_finetune/xperience10m_qwen3_omni_128ep_fullparam_pilot32_preemptible_8gpu_20260609/progress.jsonl", |
| "omni_finetune/xperience10m_qwen3_omni_128ep_fullparam_pilot32_preemptible_8gpu_20260609/launch_status.jsonl", |
| "omni_finetune/xperience10m_qwen3_omni_128ep_fullparam_pilot64_preemptible_8gpu_20260609/fullparam_pilot64_summary.json", |
| "omni_finetune/xperience10m_qwen3_omni_128ep_fullparam_pilot64_preemptible_8gpu_20260609/progress.jsonl", |
| "omni_finetune/xperience10m_qwen3_omni_128ep_fullparam_pilot64_preemptible_8gpu_20260609/training_metadata.json", |
| "omni_finetune/xperience10m_qwen3_omni_128ep_fullparam_pilot64_preemptible_8gpu_20260609/RUN_REPORT.md", |
| "omni_finetune/xperience10m_qwen3_omni_128ep_fullparam_pilot64_preemptible_8gpu_20260609/config.yaml", |
| "omni_finetune/xperience10m_qwen3_omni_128ep_fullparam_pilot64_preemptible_8gpu_20260609/launch_status.jsonl", |
| "omni_finetune/xperience10m_qwen3_omni_128ep_fullparam_pilot128_preemptible_8gpu_20260609/fullparam_pilot128_summary.json", |
| "omni_finetune/xperience10m_qwen3_omni_128ep_fullparam_pilot128_preemptible_8gpu_20260609/progress.jsonl", |
| "omni_finetune/xperience10m_qwen3_omni_128ep_fullparam_pilot128_preemptible_8gpu_20260609/launch_status.jsonl", |
| "omni_finetune/xperience10m_qwen3_omni_128ep_fullparam_pilot128_after_qwen_v5_preemptible_8gpu_20260609/progress.jsonl", |
| "omni_finetune/xperience10m_qwen3_omni_128ep_fullparam_pilot128_after_qwen_v5_preemptible_8gpu_20260609/config.yaml", |
| "omni_finetune/xperience10m_qwen3_omni_128ep_fullparam_pilot128_after_qwen_v5_preemptible_8gpu_20260609/training_metadata.json", |
| "omni_finetune/xperience10m_qwen3_omni_128ep_fullparam_pilot128_after_qwen_v5_preemptible_8gpu_20260609/RUN_REPORT.md", |
| "omni_finetune/xperience10m_qwen3_omni_128ep_fullparam_pilot128_after_qwen_v5_preemptible_8gpu_20260609/launch_status.jsonl", |
| "omni_finetune/HF_UPLOAD.md", |
| "omni_finetune/model_output_probe_readiness/model_output_probe_readiness.json", |
| "omni_finetune/model_output_probe_readiness/RUN_REPORT.md", |
| "omni_finetune/xperience10m_cosmos3_super_training_readiness_20260607/training_readiness.json", |
| "omni_finetune/xperience10m_cosmos3_super_training_readiness_20260607/RUN_REPORT.md", |
| "omni_finetune/xperience10m_cosmos3_super_training_readiness_metadata_a100_20260609/training_readiness.json", |
| "omni_finetune/xperience10m_cosmos3_super_training_readiness_metadata_a100_20260609/RUN_REPORT.md", |
| "omni_finetune/xperience10m_cosmos3_super_training_contract_audit_camera_pose_20260608/training_contract_audit.json", |
| "omni_finetune/xperience10m_cosmos3_super_training_contract_audit_camera_pose_20260608/RUN_REPORT.md", |
| "omni_finetune/xperience10m_cosmos3_super_action_packer_schema_smoke_20260608/packer_summary.json", |
| "omni_finetune/xperience10m_cosmos3_super_action_packer_schema_smoke_20260608/RUN_REPORT.md", |
| "omni_finetune/verified_public/xperience10m_cosmos3_nano_128ep_future_window_h5_compat_adapter_eval_test_full/verified_result_summary.json", |
| "omni_finetune/verified_public/xperience10m_cosmos3_nano_128ep_future_window_h5_compat_adapter_eval_test_full/PUBLIC_RESULT_SUMMARY.md", |
| "omni_finetune/verified_public/xperience10m_cosmos3_nano_128ep_future_window_h5_compat_adapter_eval_test_full/eval/metrics.json", |
| "omni_finetune/verified_public/xperience10m_cosmos3_nano_128ep_future_window_h5_compat_adapter_eval_test_full/eval/RUN_REPORT.md", |
| ] |
|
|
| DOC_FILES = [ |
| "README.zh.md", |
| "README.es.md", |
| "README.fr.md", |
| "README.de.md", |
| "README.ja.md", |
| "README.ko.md", |
| "README.pt.md", |
| "ARTIFACT_GUIDE.md", |
| "OMNI_MODEL_EXTENSION_CONTRACT.md", |
| "QUALITY_GATES.md", |
| "EVALUATION_PROTOCOL.md", |
| "FIGURE_INDEX.md", |
| "FOUNDATION_MODEL_PLAN.md", |
| "GLOSSARY.md", |
| "ADDITIONAL_DEVELOPMENT_DIRECTIONS.md", |
| "THREE_FOUNDATION_PIPELINES.md", |
| "TWO_EVIDENCE_LINES.md", |
| "TWO_EVIDENCE_LINE_RESULT_SUMMARY.md", |
| "QWEN3_OMNI_RUN_LINEAGE.md", |
| "PROJECT_BRIEF.md", |
| "PUBLIC_READER_MAP.md", |
| "RENDERED_SITE_CHECK.md", |
| "RESEARCH_ROADMAP.md", |
| "PROJECT_STATUS.md", |
| "REPRODUCIBILITY.md", |
| "TASK_SUITE_ENHANCEMENT_128.md", |
| "XPERIENCE10M_128_EPISODE_FEATURE_INDEX.md", |
| "TASK_METHOD_20_GAP_AUDIT.md", |
| "TASK_METHOD_20_RESULT_MATRIX.md", |
| "TASK_SUITE_20.md", |
| "PUBLIC_SURFACE_QA.md", |
| "EVIDENCE_CONTRACT.md", |
| "RESEARCH_TAKEAWAYS.md", |
| "SOURCE_ALIGNMENT_AUDIT.md", |
| "XPERIENCE_EMBODIED_FOUNDATION_MODEL_PRETRAINING.md", |
| "XPERIENCE10M_DATASET_CARD_ALIGNMENT.md", |
| ] |
|
|
|
|
| def sha256(path: Path) -> str: |
| digest = hashlib.sha256() |
| with path.open("rb") as handle: |
| for chunk in iter(lambda: handle.read(1024 * 1024), b""): |
| digest.update(chunk) |
| return digest.hexdigest() |
|
|
|
|
| def display_path(path: Path, hf_root: Path) -> str: |
| resolved = path.resolve() |
| bases = [ |
| ("hf_space", hf_root / "space"), |
| ("hf_artifacts", hf_root / "artifacts"), |
| ("hf_model", hf_root / "model"), |
| ("repo", ROOT), |
| ("hf_publish", hf_root), |
| ] |
| for label, base in bases: |
| try: |
| return f"{label}:{resolved.relative_to(base.resolve()).as_posix()}" |
| except ValueError: |
| continue |
| return path.name |
|
|
|
|
| def file_record(path: Path, hf_root: Path) -> dict: |
| record = { |
| "path": display_path(path, hf_root), |
| "exists": path.exists(), |
| } |
| if path.exists() and path.is_file(): |
| record["bytes"] = path.stat().st_size |
| record["sha256"] = sha256(path) |
| else: |
| record["bytes"] = 0 |
| record["sha256"] = None |
| return record |
|
|
|
|
| def verified_public_result_files() -> list[str]: |
| """Return every public-safe file from verified model packages. |
| |
| Verified packages are already sanitized by the package audit; mirroring the |
| whole package prevents final model results from being silently omitted when a |
| new run lands under results/omni_finetune/verified_public. |
| """ |
|
|
| verified_root = ROOT / "results/omni_finetune/verified_public" |
| if not verified_root.exists(): |
| return [] |
| files: list[str] = [] |
| for path in verified_root.rglob("*"): |
| if not path.is_file(): |
| continue |
| files.append(path.relative_to(ROOT / "results").as_posix()) |
| return sorted(files) |
|
|
|
|
| def tier2_result_files() -> list[str]: |
| """Return every generated public-safe artifact for tasks 13-20. |
| |
| The directory name is historical and kept for stable public links. |
| """ |
|
|
| tier2_root = ROOT / "results/episode_task_suite/tier2_task_suite" |
| if not tier2_root.exists(): |
| return [] |
| files: list[str] = [] |
| for path in tier2_root.rglob("*"): |
| if path.is_file(): |
| files.append(path.relative_to(ROOT / "results").as_posix()) |
| return sorted(files) |
|
|
|
|
| def a100_128_metadata_result_files() -> list[str]: |
| """Return the public-safe 128-episode metadata baseline rerun artifacts.""" |
|
|
| result_root = ROOT / "results/omni_finetune/a100_128_metadata_task_baselines_20260616_v2" |
| if not result_root.exists(): |
| return [] |
| files: list[str] = [] |
| for path in result_root.rglob("*"): |
| if path.is_file(): |
| files.append(path.relative_to(ROOT / "results").as_posix()) |
| return sorted(files) |
|
|
|
|
| def a100_128_raw20_result_files() -> list[str]: |
| """Return public-safe 128-episode raw-feature 20-task baseline artifacts.""" |
|
|
| result_root = ROOT / "results/omni_finetune/a100_128_raw20_task_baselines_complete20_proxy_20260616T091500Z" |
| if not result_root.exists(): |
| return [] |
| files: list[str] = [] |
| for path in result_root.rglob("*"): |
| if path.is_file(): |
| files.append(path.relative_to(ROOT / "results").as_posix()) |
| return sorted(files) |
|
|
|
|
| def xperience10m_128_data_feature_files() -> list[str]: |
| """Return selected 128-episode source and processed-feature artifacts.""" |
|
|
| explicit = [ |
| ROOT / "results/omni_finetune/xperience10m_128_episode_selection.json", |
| ROOT / "results/omni_finetune/xperience10m_128_episode_selection.csv", |
| ROOT / "results/omni_finetune/xperience10m_128_episode_download_files.txt", |
| ROOT / "results/omni_finetune/episode_manifest.json", |
| ROOT / "results/omni_finetune/dataset_manifest.json", |
| ROOT / "results/omni_finetune/multi_episode_128_task_baselines/metadata_feature_matrix.npz", |
| ROOT / "results/omni_finetune/a100_128_metadata_task_baselines_20260616_v2/metadata_feature_matrix.npz", |
| ] |
| dense_root = ROOT / "results/omni_finetune/xperience10m_128ep_dense_multiscale_hierarchical_v1_20260608" |
| files: list[str] = [] |
| for path in explicit: |
| if path.is_file(): |
| files.append(path.relative_to(ROOT / "results").as_posix()) |
| if dense_root.exists(): |
| for path in dense_root.rglob("*"): |
| if path.is_file(): |
| files.append(path.relative_to(ROOT / "results").as_posix()) |
| return sorted(set(files)) |
|
|
|
|
| def model_output_task_probe_result_files() -> list[str]: |
| """Return task-specific scores derived from existing verified model outputs.""" |
|
|
| result_root = ROOT / "results/omni_finetune/model_output_task_probes_20260616" |
| if not result_root.exists(): |
| return [] |
| files: list[str] = [] |
| for path in result_root.rglob("*"): |
| if path.is_file(): |
| files.append(path.relative_to(ROOT / "results").as_posix()) |
| return sorted(files) |
|
|
|
|
| def qwen3_future_task_probe_result_files() -> list[str]: |
| """Return public-safe Qwen3 future-task probe outputs from the shard run.""" |
|
|
| result_root = ROOT / "results/omni_finetune" / QWEN3_FUTURE_TASK_PROBE_RUN_ID |
| files: list[str] = [] |
| if result_root.exists(): |
| for path in result_root.rglob("*"): |
| if path.is_file(): |
| files.append(path.relative_to(ROOT / "results").as_posix()) |
|
|
| launcher_log = ( |
| ROOT |
| / "results/omni_finetune/deferred_launchers" |
| / f"{QWEN3_FUTURE_TASK_PROBE_RUN_ID}.launcher.log" |
| ) |
| if launcher_log.is_file(): |
| files.append(launcher_log.relative_to(ROOT / "results").as_posix()) |
| return sorted(files) |
|
|
|
|
| def qwen3_retrieval_task_probe_result_files() -> list[str]: |
| """Return public-safe Qwen3 retrieval probe outputs from task 08/09 runs.""" |
|
|
| files: list[str] = [] |
| for run_id in QWEN3_RETRIEVAL_TASK_PROBE_RUN_IDS: |
| result_root = ROOT / "results/omni_finetune" / run_id |
| if result_root.exists(): |
| for path in result_root.rglob("*"): |
| if path.is_file(): |
| files.append(path.relative_to(ROOT / "results").as_posix()) |
|
|
| launch_log = ROOT / "results/omni_finetune/deferred_launchers" / f"{run_id}.launch.log" |
| if launch_log.is_file(): |
| files.append(launch_log.relative_to(ROOT / "results").as_posix()) |
| direct_launch_log = ROOT / "results/omni_finetune/deferred_launchers" / f"{run_id}.launcher.log" |
| if direct_launch_log.is_file(): |
| files.append(direct_launch_log.relative_to(ROOT / "results").as_posix()) |
| return sorted(set(files)) |
|
|
|
|
| def cosmos3_super_retrieval_task_probe_result_files() -> list[str]: |
| """Return public-safe Cosmos3-Super retrieval probe outputs.""" |
|
|
| files: list[str] = [] |
| compact_names = {"RUN_REPORT.md", "collection_validation.json", "metrics.json", "server_info.json", "summary.json"} |
| for run_id in COSMOS3_SUPER_RETRIEVAL_TASK_PROBE_RUN_IDS: |
| result_root = ROOT / "results/omni_finetune" / run_id |
| if result_root.exists(): |
| for path in result_root.rglob("*"): |
| if path.is_file() and (path.name in compact_names or path.name.endswith(".progress.jsonl")): |
| files.append(path.relative_to(ROOT / "results").as_posix()) |
|
|
| for suffix in ("launch", "launcher", "retrieval", "runner"): |
| launch_log = ROOT / "results/omni_finetune/deferred_launchers" / f"{run_id}.{suffix}.log" |
| if launch_log.is_file(): |
| files.append(launch_log.relative_to(ROOT / "results").as_posix()) |
| return sorted(set(files)) |
|
|
|
|
| def cosmos3_super_future_task_probe_result_files() -> list[str]: |
| """Return public-safe Cosmos3-Super future-task probe outputs.""" |
|
|
| files: list[str] = [] |
| compact_names = { |
| "RUN_REPORT.md", |
| "collection_validation.json", |
| "metrics.json", |
| "per_class_metrics.csv", |
| "server_info.json", |
| "summary.json", |
| } |
| for run_id in COSMOS3_SUPER_FUTURE_TASK_PROBE_RUN_IDS: |
| result_root = ROOT / "results/omni_finetune" / run_id |
| if result_root.exists(): |
| for path in result_root.rglob("*"): |
| if path.is_file() and (path.name in compact_names or path.name.endswith(".progress.jsonl")): |
| files.append(path.relative_to(ROOT / "results").as_posix()) |
|
|
| for suffix in ("launch", "launcher", "runner"): |
| launch_log = ROOT / "results/omni_finetune/deferred_launchers" / f"{run_id}.{suffix}.log" |
| if launch_log.is_file(): |
| files.append(launch_log.relative_to(ROOT / "results").as_posix()) |
| return sorted(set(files)) |
|
|
|
|
| def cosmos3_super_interaction_text_task_probe_result_files() -> list[str]: |
| """Return public-safe Cosmos3-Super task-15 interaction-text probe outputs.""" |
|
|
| files: list[str] = [] |
| compact_names = { |
| "RUN_REPORT.md", |
| "collection_validation.json", |
| "confusion_matrix.csv", |
| "launch_env.txt", |
| "metrics.json", |
| "per_class_metrics.csv", |
| "server_info.json", |
| "summary.json", |
| } |
| for run_id in COSMOS3_SUPER_INTERACTION_TEXT_TASK_PROBE_RUN_IDS: |
| result_root = ROOT / "results/omni_finetune" / run_id |
| if result_root.exists(): |
| for path in result_root.rglob("*"): |
| if path.is_file() and (path.name in compact_names or path.name.endswith(".progress.jsonl")): |
| files.append(path.relative_to(ROOT / "results").as_posix()) |
|
|
| for suffix in ("launch", "launcher", "runner"): |
| launch_log = ROOT / "results/omni_finetune/deferred_launchers" / f"{run_id}.{suffix}.log" |
| if launch_log.is_file(): |
| files.append(launch_log.relative_to(ROOT / "results").as_posix()) |
| return sorted(set(files)) |
|
|
|
|
| def parity_group(name: str, local_path: Path, mirrors: dict[str, Path], hf_root: Path) -> dict: |
| local = file_record(local_path, hf_root) |
| mirror_records = {surface: file_record(path, hf_root) for surface, path in mirrors.items()} |
| failures = [] |
| if not local["exists"]: |
| failures.append({"surface": "repo", "kind": "missing", "path": local["path"]}) |
| for surface, record in mirror_records.items(): |
| if not record["exists"]: |
| failures.append({"surface": surface, "kind": "missing", "path": record["path"]}) |
| continue |
| if local["exists"] and record["sha256"] != local["sha256"]: |
| failures.append( |
| { |
| "surface": surface, |
| "kind": "hash_mismatch", |
| "path": record["path"], |
| "expected_sha256": local["sha256"], |
| "actual_sha256": record["sha256"], |
| } |
| ) |
| return { |
| "name": name, |
| "status": "pass" if not failures else "fail", |
| "local": local, |
| "mirrors": mirror_records, |
| "failures": failures, |
| } |
|
|
|
|
| def build_report(hf_root: Path) -> dict: |
| groups = [] |
|
|
| for filename in DATA_FILES: |
| groups.append( |
| parity_group( |
| f"data/{filename}", |
| ROOT / "docs/data" / filename, |
| { |
| "hf_space": hf_root / "space/data" / filename, |
| "hf_artifacts_data": hf_root / "artifacts/data" / filename, |
| "hf_artifacts": hf_root / "artifacts/docs/data" / filename, |
| "hf_model_data": hf_root / "model/data" / filename, |
| "hf_model_docs_data": hf_root / "model/docs/data" / filename, |
| "hf_model": hf_root / "model/metrics" / filename, |
| }, |
| hf_root, |
| ) |
| ) |
|
|
| for filename in ASSET_FILES: |
| groups.append( |
| parity_group( |
| f"assets/{filename}", |
| ROOT / "docs/assets" / filename, |
| { |
| "hf_space": hf_root / "space/assets" / filename, |
| "hf_artifacts_docs": hf_root / "artifacts/docs/assets" / filename, |
| "hf_artifacts_card": hf_root / "artifacts/assets" / filename, |
| "hf_model": hf_root / "model/assets" / filename, |
| }, |
| hf_root, |
| ) |
| ) |
|
|
| for filename in SCRIPT_FILES: |
| groups.append( |
| parity_group( |
| f"scripts/{filename}", |
| ROOT / "scripts" / filename, |
| { |
| "hf_artifacts": hf_root / "artifacts/scripts" / filename, |
| "hf_model": hf_root / "model/scripts" / filename, |
| }, |
| hf_root, |
| ) |
| ) |
|
|
| for filename in WEBSITE_FILES: |
| groups.append( |
| parity_group( |
| f"website/{filename}", |
| ROOT / "docs" / filename, |
| { |
| "hf_space": hf_root / "space" / filename, |
| "hf_artifacts_root": hf_root / "artifacts" / filename, |
| "hf_artifacts_docs": hf_root / "artifacts/docs" / filename, |
| "hf_model": hf_root / "model" / filename, |
| "hf_model_docs": hf_root / "model/docs" / filename, |
| }, |
| hf_root, |
| ) |
| ) |
|
|
| result_files = sorted( |
| set(RESULT_FILES) |
| | set(verified_public_result_files()) |
| | set(tier2_result_files()) |
| | set(a100_128_metadata_result_files()) |
| | set(a100_128_raw20_result_files()) |
| | set(xperience10m_128_data_feature_files()) |
| | set(model_output_task_probe_result_files()) |
| | set(qwen3_future_task_probe_result_files()) |
| | set(qwen3_retrieval_task_probe_result_files()) |
| | set(cosmos3_super_retrieval_task_probe_result_files()) |
| | set(cosmos3_super_future_task_probe_result_files()) |
| | set(cosmos3_super_interaction_text_task_probe_result_files()) |
| ) |
| for filename in result_files: |
| groups.append( |
| parity_group( |
| f"results/{filename}", |
| ROOT / "results" / filename, |
| { |
| "hf_artifacts": hf_root / "artifacts/results" / filename, |
| "hf_model": hf_root / "model/results" / filename, |
| }, |
| hf_root, |
| ) |
| ) |
|
|
| space_result_files = sorted( |
| set(RESULT_FILES) |
| | set(tier2_result_files()) |
| | set(model_output_task_probe_result_files()) |
| | set(qwen3_future_task_probe_result_files()) |
| | set(qwen3_retrieval_task_probe_result_files()) |
| | set(cosmos3_super_retrieval_task_probe_result_files()) |
| | set(cosmos3_super_future_task_probe_result_files()) |
| | set(cosmos3_super_interaction_text_task_probe_result_files()) |
| ) |
| for filename in space_result_files: |
| groups.append( |
| parity_group( |
| f"space_results/{filename}", |
| ROOT / "results" / filename, |
| { |
| "hf_space": hf_root / "space/results" / filename, |
| }, |
| hf_root, |
| ) |
| ) |
|
|
| for filename in DOC_FILES: |
| groups.append( |
| parity_group( |
| f"docs/{filename}", |
| ROOT / filename, |
| { |
| "hf_space": hf_root / "space" / filename, |
| "hf_artifacts": hf_root / "artifacts" / filename, |
| "hf_model": hf_root / "model" / filename, |
| }, |
| hf_root, |
| ) |
| ) |
|
|
| failures = [ |
| {"group": group["name"], **failure} |
| for group in groups |
| for failure in group["failures"] |
| ] |
| by_surface: dict[str, int] = {} |
| for failure in failures: |
| by_surface[failure["surface"]] = by_surface.get(failure["surface"], 0) + 1 |
|
|
| return { |
| "status": "pass" if not failures else "fail", |
| "generated_at_utc": datetime.now(timezone.utc).isoformat(timespec="seconds"), |
| "hf_root": "hf_publish", |
| "summary": { |
| "group_count": len(groups), |
| "failure_count": len(failures), |
| "failures_by_surface": by_surface, |
| }, |
| "checks": [ |
| { |
| "name": "repo_hf_space_artifact_model_data_parity", |
| "status": "pass" |
| if not any(failure["group"].startswith("data/") for failure in failures) |
| else "fail", |
| }, |
| { |
| "name": "repo_hf_visual_asset_parity", |
| "status": "pass" |
| if not any(failure["group"].startswith("assets/") for failure in failures) |
| else "fail", |
| }, |
| { |
| "name": "repo_hf_validator_script_parity", |
| "status": "pass" |
| if not any(failure["group"].startswith("scripts/") for failure in failures) |
| else "fail", |
| }, |
| { |
| "name": "repo_hf_website_html_parity", |
| "status": "pass" |
| if not any(failure["group"].startswith("website/") for failure in failures) |
| else "fail", |
| }, |
| { |
| "name": "repo_hf_diagnostic_result_parity", |
| "status": "pass" |
| if not any(failure["group"].startswith("results/") for failure in failures) |
| else "fail", |
| }, |
| { |
| "name": "repo_hf_quality_doc_parity", |
| "status": "pass" |
| if not any(failure["group"].startswith("docs/") for failure in failures) |
| else "fail", |
| }, |
| ], |
| "groups": groups, |
| "failures": failures, |
| } |
|
|
|
|
| def main() -> int: |
| parser = argparse.ArgumentParser() |
| parser.add_argument("--hf-root", type=Path, default=DEFAULT_HF_ROOT) |
| parser.add_argument("--output", type=Path, default=DEFAULT_OUTPUT) |
| args = parser.parse_args() |
|
|
| report = build_report(args.hf_root.resolve()) |
| args.output.parent.mkdir(parents=True, exist_ok=True) |
| args.output.write_text(json.dumps(report, indent=2) + "\n", encoding="utf-8") |
| print(f"{report['status'].upper()}: wrote {args.output}") |
| if report["status"] != "pass": |
| for failure in report["failures"][:40]: |
| print(f"- {failure['group']}: {failure['surface']} {failure['kind']} {failure['path']}") |
| if len(report["failures"]) > 40: |
| print(f"- ... {len(report['failures']) - 40} more failures") |
| return 1 |
| return 0 |
|
|
|
|
| if __name__ == "__main__": |
| raise SystemExit(main()) |
|
|
